other
0.6B parameters
Commercial OK
Reviewed May 2026

Parakeet TDT 0.6B v2

600M-parameter FastConformer-TDT transducer ASR from NVIDIA NeMo. Topped the Hugging Face Open ASR Leaderboard in 2025 for English, with WER ~6.05% averaged across the leaderboard suite. Outputs word/segment timestamps natively and handles up to 24-minute audio in one pass.

License: cc-by-4.0·Context: 0 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

If you have an NVIDIA GPU and only need English, Parakeet beats Distil-Whisper on both WER and latency. Outside that envelope, Distil-Whisper is easier to deploy.

Overview

600M-parameter FastConformer-TDT transducer ASR from NVIDIA NeMo. Topped the Hugging Face Open ASR Leaderboard in 2025 for English, with WER ~6.05% averaged across the leaderboard suite. Outputs word/segment timestamps natively and handles up to 24-minute audio in one pass.

Strengths

  • Leaderboard-leading English WER among <1B open ASR models
  • Native long-form (up to 24 min) without chunking workarounds
  • Streaming-capable transducer architecture, low latency on GPU
  • Word-level timestamps without forced alignment

Weaknesses

  • English-only — no multilingual coverage at all
  • CC-BY-4.0 requires attribution; not as frictionless as Apache/MIT for product use
  • NeMo runtime is heavier than whisper.cpp; CPU inference is impractical
  • Tied to NVIDIA toolchain — ONNX/CoreML export is community-maintained, not first-class

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M0.3 GB1 GB

Get the model

HuggingFace

Original weights

huggingface.co/nvidia/parakeet-tdt-0.6b-v2

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Parakeet TDT 0.6B v2.

Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Step down
Smaller — faster, runs on weaker hardware
No verdicted models in the next tier down yet.

Frequently asked

What's the minimum VRAM to run Parakeet TDT 0.6B v2?

1GB of VRAM is enough to run Parakeet TDT 0.6B v2 at the Q4_K_M quantization (file size 0.3 GB). Higher-quality quantizations need more.

Can I use Parakeet TDT 0.6B v2 commercially?

Yes — Parakeet TDT 0.6B v2 ships under the cc-by-4.0, which permits commercial use. Always read the license text before deployment.

What's the context length of Parakeet TDT 0.6B v2?

Parakeet TDT 0.6B v2 supports a context window of 0 tokens (about 0K).

Source: huggingface.co/nvidia/parakeet-tdt-0.6b-v2

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.

Related — keep moving

Before you buy

Verify Parakeet TDT 0.6B v2 runs on your specific hardware before committing money.